Daily space time precipitation generator (Triple_M)

►  Get access to model via GitHub

“TripleM” (Multisite Markov Model) is a scientific tool for the fast simulation of synthetic daily precipitation time series at one or multiple sites as required for many different studies in earth sciences such as atmospheric sciences, hydrology, hazard and risk assessment or climate impact assessment.

When simulating a single site, the code corresponds to a classic ‘Richardson type model’, where precipitation occurrences are first simulated with a Markov chain and precipitation amounts are randomly sampled, either from a parametric distribution or from the observations. When two or more sites are simulated, daily snapshots of precipitation occurrences at all sites – also called ‘occurrence vectors’ - are clustered into similar precipitation patterns, which are then simulated in a Markovian process.

In case of multiple sites, the model can either be used as a pure resampling (bootstrap) algorithm, meaning that the simulated precipitation is built from the observations, or in a parametric version, where the resampled observed precipitation amounts are replaced by synthetic precipitation amounts, which allows for the simulation of unobserved precipitation extremes. The model can be set up in a monthly or seasonal setup depending on the requirements of the application.